Development of a magnetically aligned regenerative tissue-engineered electronic nerve interface for peripheral nerve applications

Abstract

ABSTRACT

Peripheral nerve injuries can be debilitating to motor and sensory function, with severe cases often resulting in complete limb amputation. Over the past two decades, prosthetic limb technology has rapidly advanced to provide users with crude motor control of up to 20° of freedom; however, the nerve-interfacing technology required to provide high movement selectivity has not progressed at the same rate. The work presented here focuses on the development of a magnetically aligned regenerative tissue-engineered electronic nerve interface (MARTEENI) that combines polyimide “threads” encapsulated within a magnetically aligned hydrogel scaffold. The technology exploits tissue-engineered strategies to address concerns over traditional peripheral nerve interfaces including poor axonal sampling through the nerve and rigid substrates. A magnetically templated hydrogel is used to physically support the polyimide threads while also promoting regeneration in close proximity to the electrode sites on the polyimide. This work demonstrates the utility of magnetic templating for use in tuning the mechanical properties of hydrogel scaffolds to match the stiffness of native nerve tissue while providing an aligned substrate for Schwann cell migration in vitro. MARTEENI devices were fabricated and implanted within a 5-mm-long rat sciatic-nerve transection model to assess regeneration at 6 and 12 weeks. MARTEENI devices do not disrupt tissue remodeling and show axon densities equivalent to fresh tissue controls around the polyimide substrates. Devices are observed to have attenuated foreign-body responses around the polyimide threads. It is expected that future studies with functional MARTEENI devices will be able to record and stimulate single axons with high selectivity and low stimulation regimes.

1. Introduction

Nerve injuries can have lasting, debilitating consequences including loss of sensation, impairment of motor function, and severe pain. Approximately 200,000 peripheral nerve surgeries are performed per year in the United States alone [1]. In some cases, surgical repair of damaged nerves results in poor outcomes, with over 1600 cases from 2001 to 2015 deemed “unsalvageable” and recommended for amputation [2]. With nearly 185,000 amputations performed per year [3], over 2 million people live with limb loss in the United States alone [4].

During the past two decades, prosthetic-limb technology has rapidly advanced with artificial appendages that can provide patients with over 20° of freedom, relatively functional crude motor control, and the ability to capture some forms of sensory information (e.g., touch) [5,6]. Despite these remarkable advances in prostheses, the interfacing technology needed to comprehensively communicate bi-directionally between the user and the limb has not progressed at the same rate. Traditional prosthetic control has been restricted to external myoelectric inputs or pressure-switch controls via muscle bulging that provide limited degrees of prosthetic freedom. Additionally, myoelectric prostheses are unable to restore sensory or tactile feedback, which therein requires extensive patient training to achieve useful mastery of the limb [5,7]. To overcome these obstacles, modern engineered approaches have shifted to directly interfacing prostheses with the nervous system of amputees to engage with internal neural inputs from their amputated limb [8].

Neural interfaces implanted into the cortex have been used to capture signals for controlling external prostheses as well as to deliver signals for providing sensory feedback [9–11]. However, growing concerns over cortical-implant invasiveness prompted a technological shift to peripheral nerve interfaces, which can provide greater neural selectivity and easier clinical implementation.

Although peripheral nerve-interface technologies, including cuff [12], FINE [13], and LIFE [14,15] electrodes, can be easily implanted into the limb tissue, each approach is limited in terms of their ability to capture and stimulate axonal activity with both high spatial selectivity and over a large spatial extent (i.e., throughout the full diameter of the nerve). Transition to “regenerative” approaches (i.e., Sieve [16,17] and Regenerative Multi-Electrode Interfaces [18,19]) allow axon extension through the interface and could be scaled to provide selective access to a larger population of axons; however, these techniques employ large rigid constructs that induce inflammatory, foreign-body encapsulation causing signal decay and a reduction in the number of active sites over time [20,21]. The limitations presented by existing nerve-interfacing approaches evoke the need to develop a device that can comprehensively engage with many axons in a nerve with high selectivity and scalability, while also attenuating the foreign-body response that would further reduce device efficacy over time.

Concerns over existing neural interfaces may inspire next-generation approaches to incorporate concepts of tissue engineering and further build on the successes of the field of peripheral nerve repair. By combining regenerative approaches with neural-interface technologies, it may become possible to provide a relatively dense 3D spatial distribution of microelectrodes through the full diameter of the nerve while promoting axonal growth in close proximity to the electrodes [22]. This tissue-engineered-enhanced interfacing strategy may ultimately have the potential to increase the total population of axons from which action potentials could be selectively captured or stimulated, thereby providing greater control over functional outcomes (e.g., prosthetic movement, sensory feedback). Traditional peripheral nerve-repair strategies include “top-down” clinically available nerve guides (i.e., autografts, processed allografts, conduits). However, due to processing constraints, the use of nerve grafts limits the ability to incorporate functional microelectrodes without disruption of graft integrity [23]. Alternatively, integration of soft electronic interfaces can be achieved by immobilization within “bottom-up” biomaterial scaffolds. Tissue-engineered scaffolds have been used extensively for nerve-regeneration applications for their ability to easily incorporate biological cues and further offer an exciting opportunity to integrate nerve interfacing technology [22].

Our team has harnessed this opportunity by developing an implantable tissue-engineered electronic nerve interface (TEENI) that incorporates thin polyimide “threads” dispersed with microelectrodes and encapsulated within tissue-engineered hydrogel scaffolds (Fig. 1A) [22,24–28]. The flexible polyimide-metal microelectrode arrays are designed to interface with individual axons, while the tissue-engineered scaffold promotes neural-tissue regeneration and integration around the array. There are several advantages to this approach. First, polyimide has an elastic modulus that is approximately two orders of magnitude below traditional interfacing materials (e.g., silicon, tungsten) [29], which helps to address concerns over mechanical mismatch of the substrate and soft nerve tissue [27,30]. In addition, the mechanically tunable hydrogel scaffold is used to support and immobilize the microelectrode threads within a 3D conduit (Fig. 1B). The hydrogel scaffold primarily contains extracellular matrix (ECM)-inspired hyaluronic acid (HA) and collagen I. HA is abundantly found in peripheral nerve tissue and is upregulated after injury [31,32]. To provide mechanical tunability and fabrication control, HA was modified with photochemical cross-linkable glycidyl methacrylate (GMHA) to allow for matrix gelation around the threads. Collagen I (Col) was used to further recapitulate native nerve composition and enable cell adhesion. This combinatorial approach to nerve interfacing technology and tissue engineering addresses previous concerns of device mechanical mismatch to native tissue and poor action potential sampling throughout the nerve.

Fig. 1.

A schematic illustrating the components of a MARTEENI device. A) Design of a functional MARTEENI device implanted within a nerve defect. The device includes a hydrogel wrapped with decellularized small intestine submucosa (SIS) and electrodes embedded within a polyimide substrate that is further connected to a printed circuit board (PCB) for electrical connectivity to the device. Image reproduced with permission [28]. B) Cross-sectional view of polyimide threads encapsulated within a non-templated hydrogel. C) Cross-sectional view of polyimide threads encapsulated within a magnetically templated hydrogel with microchannels aligned in the same direction as the threads. D) Schematic of the MARTEENI hydrogel templating process using magnetic alginate microparticles (MAMs). Image modified from Ref. [33].

To fully exploit the scaffold’s bottom-up buildability, other members of our team have developed a novel magnetic templating technology that aims to recapitulate native tissue microarchitecture [33,34]. This technology uses magnetic alginate microparticles (MAMs) to create aligned, linear channels within GMHA-Col hydrogels that mimic the basal lamina. In a separate in vivo pilot study, magnetically templated hydrogels were compared to clinically relevant isograft controls and non-templated hydrogels with an identical composition to TEENI devices previously described [27]. Magnetically templated hydrogels promoted regeneration of axon densities similar to isograft controls across a 10 mm rat sciatic nerve gap after 4 weeks post-implantation. Additionally, templated hydrogels supported the regeneration of a larger number of axons and demonstrated more thorough degradation and tissue remodeling compared to non-templated hydrogels.

Therefore, in this work, these two novel technologies (i.e., TEENI and magnetic templating) were combined to create a magnetically aligned regenerative tissue-engineered electronic nerve interface (MARTEENI) (Fig. 1C and D). This interfacing technology incorporates thin polyimide threads containing microelectrodes into a regenerative, magnetically templated hydrogel to comprehensively engage with the peripheral nerve space. The introduction of anisotropic templating to the hydrogel guides axonal growth in close proximity to the polyimide threads, while providing a growth-permissive substrate for more thorough nerve regeneration and tissue remodeling. In the current study, the physical properties of new hydrogel compositions, improved upon from studies by Spearman et al. [27] and Lacko et al. [33], were characterized and the effect of mechanical properties and microarchitecture on Schwann cell migration was validated in vitro. MARTEENI devices were fabricated and tested in a 5-mm-long rat sciatic nerve transection model to assess histological outcomes including general tissue remodeling and nerve regeneration at an early and late chronic time point (i.e., 6 and 12 weeks, respectively). Unwired devices (i.e., without electrical components) were implanted for this pilot study to establish feasibility for the technology. MARTEENI devices were compared against other clinically relevant therapeutics including nerve obtained from a genetically matched animal, similar to autograft (“Isograft”), empty conduit (“Empty”), and uninjured nerve control (“Fresh”). Extracellular-matrix (ECM) remodeling and cellular infiltration throughout the scaffold were assessed in addition to the foreign-body response to MARTEENI devices in vivo. The results of this study establish feasibility for an exciting tissue-engineered approach to neural-interfacing technologies.

2. Materials and methods

2.1. Glycidyl methacrylate hyaluronic acid (GMHA) synthesis

GMHA was synthesized as previously described [35]. In brief, hyaluronic acid sodium salt [Sigma, 53747] was dissolved at 10 mg/ml in 50% v/v acetone [Fisher, A18–20] in water at room temperature. The dissolved solution was incubated with 6.7% triethylamine [Sigma, T0886] for 30 min, and subsequently reacted with 6.3% glycidyl methacrylate [Sigma, 770342] for 24 h at room temperature. The resulting product was precipitated in 20X volumetric excess of acetone and redissolved in 50 ml of water. Precipitation and redissolution was repeated once more. Methacrylated hyaluronic acid underwent dialysis against 1X phosphate-buffered saline (PBS) for 48 h, followed by dialysis against water for 24 h using a 10 kDa molecular weight cutoff dialysis cassette. The GMHA product was sterile filtered, lyophilized for 7 d, and stored at − 20 °C until use. 2.2. Iron oxide nanoparticle synthesis

2.2. Iron oxide nanoparticle synthesis

Nanoparticles were synthesized via co-precipitation according to previously published methods [36]. Briefly, 0.1 M iron (II) chloride tetrahydrate [Sigma, 236489–500G] and 0.2 M iron (III) chloride hexahydrate [Sigma, 220299–250G] were separately dissolved in deionized water that had been deoxygenized for 30 min with nitrogen. Iron solutions were sonicated for 20 min and deoxygenated for 5 min before mixing in a glass reactor. The iron solution was heated to 75 °C and continuously deoxygenated with mechanical agitation. Ammonium hydroxide [Sigma, 4126318–1L] was added to the heated solution and reacted for 1 h at 85 °C. The resulting iron oxide colloid was cooled to room temperature and centrifuged at 1500 rpm for 10 min, followed by magnetic decantation. Iron-oxide nanoparticles were peptized with 25% w/w tetramethylammonium hydroxide (TMAOH) [Fisher, A669–212] and added in a 1:1 volumetric ratio to the volume of the initial colloid suspension. An ultrasonicator with a 1-inch-diameter probe was used to disperse the nanoparticles, operating at 80% power for 30 min. The nanoparticles were centrifuged again at 1800 rpm for 10 min. A 1:1 volumetric ratio of TMAOH was added and ultrasonication was repeated. The nanoparticles were centrifuged at 2500 rpm for 10 min and the supernatant was removed. The resulting nanoparticles were transferred into a glass jar and left to air dry in a chemical hood overnight.

2.3. Magnetic alginate microparticle (MAM) fabrication

MAMs were produced as previously described using controlled microfluidic fabrication [34]. A droplet phase was prepared with a final composition of 10 mg/ml low-viscosity sodium alginate [Sigma, 71238], 6.25 mM of calcium EDTA [Fisher C79–500, Sigma EDS-100G], and 75 mg/ml synthesized iron-oxide nanoparticles. The droplet phase was sonicated for 20 min and filtered through a 0.22-μm-diameter pore size polyvinylidene fluoride filter before microfluidic production. A separate continuous phase was prepared with 1% v/v Picosurf surfactant [Dolomite, 3200214] in HFE-7500 fluorocarbon oil [Oakwood Chemical, 051243]. A fluorophilic 3D flow-focusing microfluidic chip [Dolomite, 3200515] with a 100-μm-deep etch was used to produce MAM droplets. The droplet phase and continuous phases were injected into the system using syringe pumps with flow rates of 10 μl/min and 40 μl/min, respectively. The resulting droplets were collected in a mechanically agitated fluorocarbon oil bath. Following droplet production, a solution containing 0.1% acetic acid [Sigma, 695092] in fluorocarbon oil was added at an equal volume to the droplet collection bath and mechanically stirred for 15 min to form MAMs. The fluorocarbon oil was siphoned from the droplets using a syringe, followed by three acetone washes and magnetic decanting between washes. Finally, the particles were further crosslinked in 200 mM of calcium chloride and stored in water at 4 °C until use. Large agglomerates were removed from the particle stock using a pipette. The final volume of water to create a desired MAM stock solution was determined by estimating how many particles were produced in total and subtracting the weight of the separated microparticle agglomerates.

2.4. Magnetically templated hydrogel fabrication for mechanical and in vitro assessments

Magnetically templated hydrogels were fabricated as previously described [33,34]. Hydrogels were fabricated with 10, 15, and 20 mg/ml with channels aligned perpendicular to the hydrogel surface. GMHA was dissolved in 0.3% v/v Irgacure 2959 photoinitiator (I2959) [BASF, 55047962] solution overnight at room temperature. Hydrogel solutions were subsequently completed with 2.2% v/v MAM stock in water and 1.5 mg/ml or 3.0 mg/ml Col [Corning, 354249] for in vitro or mechanical testing, respectively. MAMs and collagen were homogenously dispersed by mixing the solution on an asymmetrical mixer [Flacktek, DAC 150.1 FVZ] at 3500 rpm for 1 min. The complete hydrogel solution was loaded into a syringe and injected into 8 × 1.7 mm silicone molds (Grace Bio-labs, 664201) and 8 × 0.8 mm silicone molds [Grace Bio-Labs, 664101]. The molded hydrogel solutions were placed within a 30 mT magnetic array at 4 °C for 30 min to allow for MAM chain alignment. Finally, hydrogels were placed under a 365 nm UV light with 16–20 mW/cm² intensity for 10 min to photocrosslink the GMHA, followed by incubation at 37 °C for 35 min to allow collagen fibrillogenesis. Following hydrogel fabrication, MAM clearance was conducted by placing hydrogels in 0.1 M EDTA solution [Fisher, SS412–1] for 5 d at 37 °C on a shaker (70 rpm). EDTA solutions were changed daily until complete MAM dissolution and stored in PBS until use.

2.5. Mechanical characterization via indentation

Stress-relaxation measurements of templated and non-templated hydrogels were obtained via non-destructive bulk indentation testing using a Bruker BioSoft In Situ Indenter. Tests were performed by indenting 6.5% of the total sample height at a rate of 20 μm/s. The probe was held at the maximum indent depth to obtain stress relaxation data; hold times varied between 10 and 30 s to allow samples to reach a quasi-static state. All tests were performed with a 3-mm-diameter spherical glass tip. Samples were kept hydrated and tested without submersion. Three locations were tested per sample, with each experimental group containing six samples (n = 6). As described by Stewart et al. [37], the Hertz contact model was used to obtain the transient modulus from force-displacement data, where the relaxation data were fit to the standard linear solid model to obtain the rate-dependent instantaneous modulus and the steady-state modulus.

2.6. Visualization of microchannels after MAM dissolution

Magnetically templated hydrogels were fabricated as described in 2.4 by adding 4 μM fluorescein-o-methacrylate [Sigma, 568864] to the hydrogel precursor solution. Images of magnetically templated hydrogels before and after MAM dissolution were acquired at 10× magnification on a Zeiss LSM 880 laser-scanning confocal microscope with 500 μm z-stacks (10 μm intervals). Channel size was quantified using ImageJ analysis.

Visualization of microchannels within MARTEENI was achieved by incubating devices in 1 mg/ml tetramethylrhodamine-labeled dextran [Invitrogen, D7139] in 8 M HCl for 1.5 h, followed by 2 × 30 min washes in 1X PBS. Labeled MARTEENI devices were imaged on a Zeiss 880 laser-scanning confocal microscope at 10× magnification, where polyimide and microchannels were visualized at 488 and 555 nm, respectively.

2.7. In vitro schwann cell migration assay into templated hydrogels

Hydrogels were prepared and trimmed using a 6-mm-diameter biopsy punch [Integra, 3336] to fit the size of a 24-well Transwell® insert with a 0.4 μm² membrane pore size [Corning, CLS3470]. Hydrogels underwent 5 d of EDTA treatment, followed by equilibration in 1X PBS for 3 d and complete Schwann cell media for 1 d (10% fetal bovine serum [Life Technologies, 10438026], 1% penicillin-streptomycin-amphotericin B [MP Biomedicals, 091674049], 20 μg/ml bovine pituitary extract [Gibco, 13028014], 4 μM forskolin [Sigma, F3917], and 10 ng/ml fibroblast growth factor [Gibco, PHG0264] in Dulbecco’s Modified Eagle’s medium [Corning, MT10013CV]).

A chemotactic gradient was established with the top Transwell® insert containing 0 ng/ml nerve growth factor (NGF) and the bottom well containing 50 ng/ml NGF [R&D Systems, 556-NG]. Passage 3 rat Schwann cells (ScienCell, R1700) were seeded on hydrogel samples at 20,000 cells/cm². Schwann cells were cultured for 7 d, with full media exchanges every 2 d.

The cultured hydrogels were fixed in 2% paraformaldehyde [Fisher, AAJ19943K2] for 45 min at room temperature, followed by 3 × 1 h washes with 1X PBS. Hydrogels were blocked for 1 h in blocking buffer containing 3% goat serum [Bio-Techne, NBP2–23475] and 0.3% Triton X-100 [Fisher, BP151–100] in 1X PBS. Next, Schwann cells were labeled with (1:500 dilution) rabbit anti-S100 primary antibody [Sigma, S2644] (Table 1) in blocking buffer for 48 h at 4 °C. Samples were washed with 0.05% v/v Tween-20 [Fisher, BP337] in 1X PBS 3 × 2 h, followed by secondary staining with (1:400 dilution) AlexaFluor 568 anti-rabbit [Invitrogen, A11011] in blocking buffer for 24 h at 4 °C. Samples were washed with 0.05% Tween in PBS 3 × 2 h, incubated with (1:1000) Hoescht 33342 [Thermo Scientific, 62249] for 10 min, and stored in 1X PBS until imaging.

Table 1.

Antibody product information used for in vitro and in vivo immunohistochemical analyses.

Antibody Type	Name	Dilution Factor	Product Info
Primary	Rabbit anti-S100	1:500	Sigma, S2644
Primary	Rabbit anti-laminin	1:500	Sigma, L9393
Primary	Mouse anti-collagen I	1:500	Sigma, C2456
Primary	Mouse anti-neurofilament heavy	1:250	DSHB Hybridoma Product, RT97
Secondary	AlexaFluor 568 anti-Rabbit	1:250	Invitrogen, A11011
Secondary	AlexaFluor 647 anti-Mouse	1:250	Invitrogen, A32728

Images were acquired at 20× magnification on a Zeiss LSM 880 laser-scanning confocal microscope with 500 μm z-stacks (10 μm intervals). Whole Transwell® inserts containing the sample were placed in a 35-mm-diameter glass-bottom dish and hydrogels were imaged through the Transwell® membrane. The number of cells per depth was determined by taking the sum of all DAPI-stained pixels per z-stack and dividing by 100 pixels, which was determined to be the average nuclei size.

2.8. Polyimide thread design

Non-functional threads (i.e., polyimide with no conductive electrodes) were designed according to prior published methods [24]. The region of each polyimide thread-set to be encapsulated in tissue-engineered hydrogel [27] consists of four individual 80-μm-wide, 10-μm-thick, and 7-mm-long “threads” spaced 160 μm apart. The individual threads were stabilized by three integrated polyimide cross-braces spaced 1 mm apart along the thread set (Supplemental Fig. 1).

2.9. Polyimide-thread fabrication

All microfabrication steps were performed in a class 100/1000 cleanroom at the University of Florida Research Service Center using 100 mm Si wafer [University Wafer, test-grade, P-type, <100>] carrier substrates. 3,3^′,4,4^′-biphenyltetracarboxilic dianhydride-p-phenylene diamine (BPDA-PDA) polyimide precursor [U Varnish S, UBE Ind.] was spin-coated at 2000 rpm and thermally cured on a hotplate [Wenesco, Inc., HP1212YH with insulated hood] in an N₂ atmosphere to 450 °C. Cured polyimide thickness of 10 μm was verified using spectral reflectance [Filmetrics, Inc., F40]. Next, a 25-μm-thick layer of photoresist [MicroChemicals GmbH, AZ9260] was spin-coated, rehydrated for 1 h at 50% humidity, exposed at 1075 mJ through a FeO-quartz photomask using a mask aligner [SUSS MicroTec, MA6], and developed in 4:1 DIW: developer [MicroChemicals GmbH, AZ400K]. Non-masked full-thickness polyimide was dry-etched using a reactive ion etch O₂-based plasma [Plasma-Therm, LLC., Unaxis 790]. Residual photoresist was removed after etching using PRS3000 [J.T. Baker®] at 70 °C. Structured thread-sets were manually removed from wafers using tweezers in preparation for integration into the hydrogel scaffold.

2.10. MARTEENI-device fabrication for in vivo implantation

The fabrication of the MARTEENI device was adapted from previous methods [27,33]. Polyimide thread-sets were ethylene oxide sterilized prior to hydrogel encapsulation and implant fabrication. Thread-sets were secured within a 3D-printed assembly jig with 1-mm-diameter dowel pins [Grainger, 38DT36] and threaded through a slit within a 1/16” diameter Tygon® tube [Fisher 14–171-129] trimmed to 5 mm in length [27]. Precursor hydrogel solutions were made under sterile conditions with 10 mg/ml GMHA, 0.3% I2959, 3.0 mg/ml collagen I, and 2.2% v/v MAMs. The solution was pipetted into the openings of the Tygon® tube mold, encapsulating the threads (Fig. 2), and the MAMs were aligned within the magnetic array followed by subsequent hydrogel crosslinking via UV photopolymerization and thermal incubation at 37 °C. MAMs were removed using a 0.1 M EDTA solution at 37 °C on a shaker (70 rpm) for 5 d, resulting in a templated hydrogel scaffold. To enable surgical handling and implantation, the molded hydrogel was removed from the Tygon® mold and wrapped with decellularized porcine small intestine submucosa (SIS) [Cook Biotech]. The SIS was cut to 7 mm × 7 mm. Four non-continuous surgical knots formed with 9–0 nylon sutures [ARO Surgical, T5A09N10] were used to form the conduit: two suture knots to secure the polyimide threads to the SIS and two to close the ends of the conduit. Finally, the device was liberated from the assembly jig by cutting the sacrificial ends of the secured polyimide threads to the length of the SIS wrap using microsurgical scissors. Finalized devices were stored in PBS at 4 °C until implantation.

Fig. 2.

MARTEENI implant fabrication. Non-functional polyimide thread-sets were secured within a Tygon® tube. 1) Hydrogel precursor solution was injected into the tube and MAMs were 2) aligned within a magnetic field; 3) MAM dissolution was achieved with EDTA to leave behind templated microchannels. Finally, the device was 4) wrapped in a decellularized small intestine submucosa wrap and polyimide threads were cut away, yielding a suturable device for implantation within a 5 mm rat sciatic nerve gap. Scale bar = 1 mm.

2.11. Surgical procedures

All animal work was performed with prior approval and in accordance with the University of Florida Institutional Animal Care and Use Committee (IACUC) guidelines (Protocol #201809095).

The MARTEENI devices (n = 6) were assessed against clinically relevant nerve guides including isograft (n = 4), empty conduit (n = 3–4), and a fresh nerve explant control (n = 3–4). Lewis rats (8 weeks old, 200–250 g [Charles River]) were randomly assigned to experimental groups using a random number generator.

Empty SIS controls were fabricated without polyimide thread-sets or hydrogel scaffolds and were sutured with three non-continuous knots (9–0 nylon sutures) to form conduits. Each conduit length was 7 mm, including 1-mm-long overhangs on each end for implantation. For the isograft group, a 5-mm-long section of sciatic nerve was harvested from a genetically-matched donor rat for transplantation into the experimental animal as described below.

Rats were anesthetized with isoflurane and core body temperature was carefully monitored through the duration of the procedure [38]. Using aseptic technique, the sciatic nerve was isolated using a longitudinal skin incision down the midline of the posterior thigh and gluteal muscle-splitting approaches. A proximal nerve transection was made 4 mm distal to the ligaments of the greater sciatic foramen and the proximal nerve stump was sutured to the 1-mm-long SIS overhangs beyond the end of the MARTEENI device. Next, the nerve was transected to remove a 3-mm-long section of the distal sciatic nerve, and the distal nerve stump was sutured to the remaining SIS overhang to ensure tension-free repair. The gluteal muscles were reattached, and the skin closed using 5–0 PGCL sutures [S-Q518R13]. After 6 or 12 weeks of repair, animals were euthanized. The regenerated devices were collected, and the contralateral nerves were also collected as fresh explant controls.

2.12. Immunohistochemical analysis

The perfused and harvested samples were immediately fixed in 4% paraformaldehyde for 24 h, followed by 1X PBS washes with buffer changes once daily for 3 d, and stored in 30% sucrose [ThermoFisher, S550] and 0.5% sodium azide [Sigma, S2002] in 1X PBS at 4 °C for at least 24 h before cryosectioning. All samples were embedded in optimal-cutting-temperature compound [Electron Microscopy Sciences, 6255001] and cross-sectioned (10 μm) with a Leica CM1950 cryostat at the midgraft, and 1 mm into the proximal and distal ends of the implant. Sections were deposited onto gelatin-coated glass slides and dried overnight. Slides were stored at − 80 °C until subsequent staining.

Slides were removed from − 80 °C and warmed at 37 °C for 2 h on a slide warmer. Slides were blocked for 1 h with blocking buffer containing 3% goat serum and 0.3% Triton X-100. Slides were incubated with primary antibodies (Table 1) rabbit anti-S100 (Schwann cells) and mouse anti-collagen I or rabbit anti-laminin and mouse anti-neurofilament heavy chain in blocking buffer for 24 h at 4 °C. Slides underwent 3 × 10 min 1X PBS washes at room temperature, followed by secondary antibody incubation with (1:250) AlexaFluor 647 anti-mouse and AlexaFluor 568 anti-rabbit in blocking buffer for 2 h at room temperature. Slides underwent 3 × 10 min 1X PBS washes at room temperature, followed by incubation with (1:1000) DAPI [ThermoFisher, D1306] in water for 10 min, followed by a 10 min 1X PBS wash at room temperature. Finally, slides were coverslipped with Fluoromount-G [SourthernBioTech, 0100–01] and clear polished to seal the coverslip edges. Fluorescence imaging was conducted on a Zeiss Axioimager Z2 and Zeiss LSM 880 laser-scanning confocal microscope for higher resolution images. Five consecutive sections were imaged for each sample at each respective implant location (i.e., proximal, midgraft, distal). Staining and imaging procedures were repeated for all samples to acquire two technical replicates.

Stained and imaged nerve sections were analyzed using ZenPro software. All performed image analyses were blinded. Regions of interest (ROIs) were analyzed for (1) axon density, (2) axon diameter, (3) collagen I average intensity, (4) laminin average intensity, (5) Schwann average intensity, and (6) foreign-body-capsule thickness (MARTEENI samples only). For each type of analysis, at least ten ROIs were quantified and averaged together for each sample and nerve location.

To quantify average axon density, the total axon count of an ROI was divided by the ROI area (mm²). To derive the average axon diameter, axons were assumed to be circular in cross-section, where average diameter (d) was calculated from the average axon area (A) value (A = πr ², where r = d/2). Collagen I, laminin, and Schwann cell average intensity were calculated from the “Mean Value Intensity” Zen software functions. Lastly, foreign-body-capsule thickness around the polyimide threads was measured from the outer edge of each thread to the outer edge of the fibrous capsule. Reported capsule measurements were averaged from four respective directions.

2.13. Statistical analysis

All statistical analyses were performed using JMP Pro 14 and R-Studio. Statistical differences between experimental groups were determined with 2-way ANOVA for mechanical and in vivo intensity measurements (i.e., S100, collagen, laminin), 3-way ANOVA for in vivo axon density and fibrotic capsule measurements, and 1-way ANOVA for axon diameter. In vitro migration data were initially plotted as histograms to examine their distribution and both Poisson and negative binomial distributions were identified as suitable representations for the counts of cells. Stepwise selection was used to compare potentially predictive models, while their fits were compared using likelihood-ratio tests and the Akaike information criterion in R-Studio [39]. A 2-way generalized linear mixed model following the negative binomial distribution was identified to best predict cell counts, especially when concentration of the hydrogel, templating process, and their interactions were considered as “fixed effects”, whereas discrete depth locations were assumed as “random effects”. To further examine the levels of discrete depth bins every 50 μm, data from templated gels only were fit to a 2-way generalized linear mixed model. Hydrogel concentration and depth were inputted as “fixed effects”, while sample levels were inputted as “random effects”. Analyses were followed by Tukey’s honestly significant difference posthoc test for multiple comparisons, with an overall confidence interval of 95%. In vitro migration data are graphed as minimum observation, lower 25% quartile (Q1), median, upper 75% quartile (Q3), and maximum observation. All other data are represented as the mean ± standard deviation (SD).

3. Results

3.1. Physical characterization of magnetically templated hydrogels

The mechanical properties of non-templated and magnetically templated hydrogels were determined via quasi-static indentation measurements (Fig. 3A). Non-templated hydrogels show significant increasing trends in both the instantaneous and steady-state moduli with respect to increasing GMHA concentration (i.e., 10, 15, 20 mg/ml) (n = 6). All concentrations of GMHA show trends of decreasing stiffness after magnetic templating (n = 6), with significant decreases observed in 15 and 20 mg/ml GMHA hydrogels. The instantaneous and steady-state moduli of freshly explanted rat sciatic nerve tissue (n = 4) were also determined to be 4.31 ± 1.29 kPa [33] and 1.31 ± 0.33 kPa, respectively. The mechanical properties of the 10 mg/mL templated hydrogel were matched to native tissue, with no statistically significant differences between the two.

Fig. 3.

Evaluation of magnetically templated hydrogel physical properties A) Quasistatic mechanical measurements demonstrate decreased instantaneous and steady-state moduli [33] following magnetic templating at all GMHA concentrations (n = 6). Green bar depicts mechanical stiffness range for fresh rat sciatic nerve tissue. Orthogonal projections of 10 mg/ml GMHA-Col templated hydrogels B) after hydrogel fabrication, before MAM dissolution and C) after MAM dissolution, demonstrating open channels at very low hydrogel stiffnesses. D) MARTEENI device visualized with polyimide threads (green) surrounded by open microchannels backfilled with fluorescent dextran (red). Scale bar = 200 μm. 2-way ANOVA, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

To visualize scaffold microarchitecture after templating at low polymer concentrations (i.e., 10 mg/ml GMHA), the hydrogel matrix was fluorescently labeled and imaged before and after MAM dissolution. Immediately after crosslinking, MAMs were visualized within the scaffold (Fig. 3B). Orthogonal projections both in the x-z and y-z planes show good alignment of MAM chains that span through the bulk of the hydrogel. After MAM dissolution, the remaining channels appear to shrink in size from 79 ± 5 to 64 ± 6 μm (p < 0.0001, unpaired t-test), but still maintain an open porosity through the hydrogel (Fig. 3C). Integration of the polyimide substrate does not inhibit alignment or degradation of MAMs, as open microchannels can still be visualized around the threads. It is noted that microchannel morphology appears different between Fig. 3B and C compared to Fig. 3D, as a result of imaging from different planes. Z-stack imaging parallel to the plane of MAM alignment, results in an image with smooth tubular microchannels; however, the true morphology of the channels after MAM dissolution is best represented by imaging perpendicular to the plane of alignment, where the channels appear similar to a string of pearls.

3.2. In vitro schwann cell migration into magnetically templated hydrogels

Maximum projections of the hydrogel surface show cell spreading on the surfaces of all hydrogel compositions (Fig. 4A). Cells were also visualized having migrated into the bulk of 10 and 15 mg/ml templated samples. No Schwann cells were visualized within 20 mg/ml hydrogels beyond the first 100 μm. Z-stack projections into the bulk of templated hydrogels show cellular infiltration of Schwann cells through the entire 500 μm z-stack depth of 10 mg/ml GMHA channels, indicated by nuclei staining. Pairwise comparisons of the average number of cells visualized through the depth of templated hydrogels (n = 5) showed significant differences between 20 mg/ml (0.1 ± 0.48 cells) and both 10 (4.47 ± 6.28 cells) and 15 mg/ml (0.43 ± 1.01 cells) (Fig. 4B). Additionally, there were no differences determined between non-templated groups, as no cells were visualized beneath the surface. Following further evaluation of cellular infiltration into discrete depth ranges within templated groups, different cell averages were detected between 10 mg/ml (3.4 ± 3.13 cells) and 15 mg/ml (0.2 ± 0.4 cells) at the farthest depth, 450–500 μm, with a confidence level greater than 94% (p = 0.0534) (Fig. 4C).

Fig. 4.

A) Representative maximum surface (top) and z-stack (bottom) projections of Schwann cell migration into magnetically templated hydrogels with varying GMHA concentrations (n = 5). Z-stack arrow + dotted line denotes maximum depth of nuclear infiltration observed within sample. Images acquired at 10× magnification. B) Box plots illustrate the distribution of cells through the depth of templated and non-templated hydrogels C) Distribution of total number of cells within templated and non-templated samples show significance between 10 mg/ml GMHA templated hydrogels vs 15 mg/ml and 20 mg/ml templated. All data is graphed as the minimum observation, lower 25% quartile (Q1), median, upper 75% quartile (Q3), and maximum observation. * indicate sample outliers at respective depths.

3.3. Schwann cell infiltration into MARTEENI and clinically relevant guides

At 6 weeks following implantation, all experimental groups exhibited Schwann cell marker, S100, expression at levels similar to Fresh (3562.6 ± 475.7 a.u.) at their midgraft, except for Empty (5836.5 ± 329.9 a.u.), which showed significantly higher S100 expression to all other groups at the earlier time point (Fig. 5). There were no differences observed in S100 expression between Isograft (4260.2 ± 1391.6 a.u.), and MARTEENI (4154.1 ± 352.9 a.u.) implants at 6 weeks. Trends of increasing S100 expression were observed at 12 weeks in all experimental groups and all groups were significantly different to Fresh (4103.3 ± 897.4 a.u.) at the midgraft. Overall, S100 expression into MARTEENI devices was equal to or greater than Fresh at 6 and 12 weeks (5315.5 ± 955.8 a.u.), respectively.

Fig. 5.

Representative fluorescence images (20× magnification) and quantitative analysis of S100 expression at the midgraft at 6 and 12 weeks. Empty had significantly higher S100 expression compared to all other groups at 6 weeks, with trends of increasing S100 expression in all experimental groups at 12 weeks. Scale bars = 10 μm. 2-way ANOVA, Tukey’s multiple comparisons, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, #p < 0.01 compared to all 6-week groups.

3.4. Temporal changes in ECM remodeling after chronic injury

To quantify changes in ECM protein expression at early and late time points, intensity measurements were obtained from immunohistochemically stained samples. At 6 weeks post-injury, significantly higher collagen I intensity is noted at the midgraft of both MARTEENI (3336.2 ± 842.5 a.u.) and Empty (3665.5 ± 810.3 a.u.) compared to Fresh (851.8 ± 363.6 a.u.) and Isograft (1250.1 ± 481.4 a.u.) (Fig. 6A). Further, at 12 weeks, MARTEENI (2116.5 ± 955.8) and Empty groups (1954.3 ± 98.6 a.u.) exhibited subsequent trends of decreasing collagen I expression to levels similar to Fresh (1012.9 ± 254.2 a.u.) and Isograft (1390.5 ± 166.9 a.u.).

Fig. 6.

Representative fluorescence images (20× magnification) and quantitative analyses of ECM proteins A) collagen I and B) laminin at the midgraft at 6 and 12 weeks. Significantly higher collagen I intensities were observed at 6 weeks in MARTEENI and Empty groups compared to Fresh, which decreases by 12 weeks. MARTEENI and Empty groups also showed laminin deposition at 6 weeks with higher intensities observed at 12 weeks compared to Fresh. These suggest differences in the roles of respective ECM proteins, where collagen is primarily deposited during the early chronic phase and laminin is deposited during the late chronic phase. Scale bars = 10 μm. 2-way ANOVA, Tukey’s multiple comparisons, p < 0.05.

At 6 weeks post-implantation, both MARTEENI (6802.8 ± 1299.8) and Empty (7045.6 ± 788.4 a.u.) groups showed elevated laminin expression compared to Fresh (2281.8 ± 1954.3 a.u.) and Isograft (4135.6 ± 2352.5 a.u.) (Fig. 6B). Trends of increasing laminin expression were observed in all experimental groups, including MARTEENI (8865.7 ± 643.3), Empty (9127.8 ± 951.8), and Isograft (4792.5 ± 2297.5 a.u.) at 12 weeks compared to 6 weeks. Differences in intensity between all experimental groups and Fresh (1588.5 ± 574.7) were statistically significant at 12 weeks.

3.5. Nerve regeneration through MARTEENI and clinically relevant guides

To obtain a holistic understanding of the regenerated tissue, explanted groups were analyzed across three regions of the nerve: proximal, midgraft, and distal. At 6 weeks, in the transection model, there were no observable differences across groups or implant regions compared to Fresh controls (Fig. 7A). At the 12-week distal implant, higher axon densities were observed in the MARTEENI (19759 ± 3269.1 a.u.) and Empty (22920 ± 3831.6 a.u.) groups compared to Fresh (15878.1 ± 18737.2 a.u.) and Isograft (18737.2 ± 1757.4 a.u.). In addition to higher axon densities in the late distal implant, experimental groups including Isograft (2.69 ± 0.24) also have significantly smaller axon diameters compared to Fresh (3.75 ± 0.12 a.u.). (Fig. 7B).

Fig. 7.

Axon density trends through the entire length of the implant for each of the experimental groups at 6 and 12 weeks. A) At 6 weeks, there were no differences observed across groups or implant regions. At 12 weeks, higher axon densities were observed in the MARTEENI and Empty groups, particularly in the distal implant. 3-way ANOVA, Tukey’s multiple comparisons, p < 0.05. B) Axon diameter at the distal stump is significantly smaller in experimental groups compared to Fresh at 12 weeks. 1-way ANOVA, p < 0.0001. C) 12-week relationships between regenerative markers, S100 vs. laminin (left) and axon density vs. laminin (right). Positive correlations were observed between regenerative markers; S100 vs laminin R2 = 0.1954, axon density vs. laminin R2 = 0.4692. Additive parallel regressions were plotted where x = laminin intensity, t = 12 week (1 if 12, 0 if 6), f = fresh (1 if fresh), g = isograft (1 if isograft), m = MARTEENI (1 if MARTEENI); empty if g/m all zero.

Correlations between regenerative markers were plotted to examine patterns of healthy regeneration in experimental groups (Fig. 7C). A positive correlation between S100 and laminin is indicated by an R²_adjusted = 0.1954. Additionally, a positive correlation between axon density and laminin is indicated by an R² _adjusted = 0.4692.

3.6. Residual artifacts of MARTEENI devices

Through qualitative image analysis, some artifacts have been observed in MARTEENI devices that may limit functional outcomes. Although MARTEENI promoted good axon densities throughout the length of the implant, some samples were shown to have residual GMHA-Col hydrogel remaining in the implant space at 6 and 12 weeks, characterized by dense, amorphous regions of collagen staining and little to no nuclear infiltration to the region (Fig. 8). It can be reasonably assumed that these were regions of remaining hydrogel because of the elevated collagen I intensity measured in these regions, as collagen I is a primary component of the GMHA-Col hydrogels used in the MARTEENI constructs.

Fig. 8.

Residual hydrogel observed within the implant space of MARTEENI samples at 12 weeks marked by dense, irregular regions of collagen I with little to no cell infiltration (S100, DAPI). Images acquired at 25× magnification (left), with regions of interest acquired at 63× magnification and separated into respective channels (right). Scale bars = 20 μm.

In addition to residual hydrogel remaining within the implant space, MARTEENI devices also experienced fibrotic encapsulation of the polyimide threads. This response was found uniformly across all MARTEENI samples with individual capsule thicknesses of approximately 20 μm across all regions of the implant and across both 6 and 12 weeks (Fig. 9).

Fig. 9.

Representative images of MARTEENI device at midgraft, 12 weeks post implantation. Dashed lines denote fibrotic capsules. 20× magnification images of A) whole implant section, scale bar = 200 μm and B) single polyimide thread with fibrotic encapsulation, scale bar = 20 μm. Thickness measurements were taken from the outermost edge of the substrate to the outermost edge of the capsule C) Quantification of capsule thickness across regions at 6 and 12 weeks show no statistical differences between implant location or timepoint.

4. Discussion

This study reports a novel magnetically aligned regenerative tissue-engineered electronic nerve interface (MARTEENI) that incorporates flexible polyimide arrays for peripheral nerve interfacing with a regenerative templated hydrogel. MARTEENI is an integrative technology that addresses previous concerns over other rigid nerve-interfacing techniques (e.g., sieve electrodes) by implementing thin, flexible polyimide electrode substrates that are two orders of magnitude lower in stiffness than traditional silicon substrates [29]. Additionally, MARTEENI provides the future capability to comprehensively interface with a larger axonal population through stacking polyimide thread sets into a GMHA-collagen hydrogel scaffold. Many other interfacing technologies, including cuff [12], FINE [13], and LIFE [14,15] electrodes are not spatially distributed through the entire tissue geometry and are thus limited in terms of the total number of individual nerve-fiber interactions. In contrast, MARTEENI provides the potential for 3D spatial distribution of microelectrodes that can be integrated directly within the nerve space to record or stimulate in a plane parallel to the general direction of axonal regeneration. Hydrogel encapsulation of the polyimide threads also provides mechanical support to the overall device. In addition, the introduction of magnetic templating into the hydrogel provides a basal lamina-mimicking substrate for Schwann cell proliferation and regenerating axons in close proximity to the electrode sites. This is expected to ultimately provide greater selectivity of single-unit axon signals. MARTEENI devices can be easily assembled by adapting a magnetic templating technique to include the suspension of polyimide threads centrally within a cylindrical mold (Fig. 2). The polyimide threads do not inhibit MAM alignment within the mold. After hydrogel fabrication and post-processing MAM removal, the threads remain immobilized within the hydrogel and can be easily wrapped within an outer layer of SIS membrane for easy implantation.

The work presented builds off of prior work by Spearman et al. [27] and Lacko et al. [33] to determine a hydrogel composition that can be used to integrate the two respective technologies. A major criterion for MARTEENI scaffold design is ensuring mechanical robustness that can support both thread integration and the templating process, while optimizing the composition to promote tissue regeneration. In addition, previous studies have suggested that modulating scaffold stiffness can direct cell behavior via mechanotransduction [40–44]. For instance, dorsal root ganglia explants demonstrate enhanced neurite outgrowth on softer material substrates that more closely resemble neural tissue [41–43]. Therefore, it is hypothesized that a scaffold with matched stiffnesses to native neural tissue would better direct cell responses such as Schwann cell migration and axon regeneration. Magnetic templating [23,26] can be used as a tool to systematically tune the mechanical properties of hydrogel scaffolds, while simultaneously introducing anisotropic microarchitectures that mimic the native structure of nerve tissue. Mechanical characterization of GMHA-Col scaffolds shows that by increasing the GMHA concentration alone, there is a predictable increase in scaffold mechanical stiffness (Fig. 3A); however, the introduction of magnetic templating resulted in dramatic decreases in stiffness compared to non-templated hydrogels. These results can be attributed to the increased void fraction within templated samples. Importantly, after templating, there are no significant differences between the stiffness of 10 mg/ml templated hydrogels and fresh rat sciatic nerve tissue, indicating matched mechanical properties and similar viscoelasticity to native tissue.

Following mechanical characterization, magnetically templated hydrogels with 10 mg/ml GMHA were fluorescently labeled and imaged to determine if low concentrations of GMHA can support the templating process and retain channel structure after MAM processing. In Fig. 3B, after templating with 10 mg/ml GMHA, MAM chains can be visualized through the scaffold depth. Weaker mechanical properties resulting from MAM dissolution appear to cause some geometric changes to the microchannels after dissolution (i.e., diameter shrinkage and shorter length channels). These microarchitectural changes suggest that 10 mg/ ml GMHA hydrogels are at the lower limit of our ability to template, however the channels still appear open and do not collapse after MAM dissolution. These results demonstrate the ability to successfully template hydrogels with soft mechanical properties without compromising microarchitectural integrity.

Finally, to evaluate the effect of scaffold stiffness and microarchitecture on modulating cell behaviors, in vitro Schwann cell migration into magnetically templated channels was assessed using hydrogel compositions that were mechanically characterized. Schwann cells facilitate nerve regeneration by secreting neurotrophins [45] and depositing ECM [46] in vivo, thus, evaluating Schwann cell migration in vitro was of particular interest before proceeding with animal experiments. Schwann cells were visualized on the surface of GMHA-Col hydrogels in both templated (Fig. 4) and non-templated groups (Supplemental Fig. 3), however, in vitro migration into the bulk of hydrogel samples was only observed in templated hydrogels. No cells were visualized beneath the surface of non-templated hydrogels at any composition, emphasizing the necessary role of anisotropy on Schwann cell infiltration. Moreover, in vitro migration was only observed in templated groups with stiffnesses that were most similar to fresh nerve tissue (i.e., 10 and 15 mg/ml). Further quantification of the average number of cells at specific channel depths found that 10 mg/ml templated hydrogels promoted infiltration through the furthest channel depth, compared to 15 mg/ml. These data support previous findings that suggest that bulk mechanical properties [47,48] and 3D structure [49,50] individually play roles in Schwann cell behavior in vitro, including elongation, migration, and localization with neuronal cells. Specifically, these data show that in vitro Schwann cell migration is better promoted by scaffolds with combined mechanical properties and microarchitecture that mimic fresh nerve tissue. Thus, 10 mg/ml GMHA was chosen as the main component of our MARTEENI devices, with the inclusion of 3 mg/ml Col, 0.3% I2959, and 2.2% MAMs for subsequent in vivo studies.

A pilot in vivo study was performed to assess general tissue remodeling within MARTEENI devices. Quantitative assessments were compared to clinically relevant controls to understand how devices affect the regeneration process. First, Schwann cell expression of S100 [51] was evaluated at two chronic time points after implantation using immunohistochemical analysis (Fig. 5). At 6 weeks, a statistically significant increase in S100 expression within Empty conduits is observed, which can be attributed to the open lumen of the conduit and the lack of intraluminal filler that would otherwise slow Schwann cell infiltration at short defect lengths. Trends of increasing S100 expression were seen at the late time point in all experimental groups. These delayed observations of increased S100 fluorescence intensity at 12 weeks could be attributed to luminal clearance of donor cellular debris [52,53], as well as scaffold matrix in Isograft and MARTEENI, respectively [54]. Although Schwann cells contribute to intraluminal turnover, including clearance of their own myelin sheaths, other cell types such as macrophages play larger roles in matrix clearance. This may also support lower S100 expression at 6 weeks in Isograft and MARTEENI [55,56]. Overall, increased S100 expression is observed in all experimental groups, which is indicative of normal and expected Schwann cell migration into the site of injury to mediate signaling for early stages of nerve regeneration.

Schwann cell presence through the various stages of peripheral nerve repair is critical to the success of axonal elongation, as Schwann cells mediate nascent extracellular-matrix deposition [46]. Specifically, Schwann cells synthesize the basal lamina that guide regenerating axons [57,58]. At 6 weeks post-injury, significantly higher collagen I and laminin deposition were noted at the midgraft in both MARTEENI and Empty groups (Fig. 6). The initial increase in ECM protein expression could be attributed to matrix stabilization [59]. As cells infiltrate into non-native guides, there is a need to rapidly deposit extracellular proteins into the space to stabilize the matrix for successful axonal elongation [60,61]. Interestingly, at 12 weeks, MARTEENI and Empty groups exhibited trends of decreasing collagen I expression and an overall increase in laminin expression. The deviation of trends from 6 to 12 weeks suggests that the injury site has reached a chronic state where new matrix has been deposited and has transitioned primarily to remodeling [61]. These observations support that different ECM proteins mediate regeneration in specialized ways. Collagen I serves as mechanical support to stabilize epineurial and endoneurial structures following acute insult to the tissue [62,63]. Laminin, which comprises a major portion of the basal lamina, is not only secreted by Schwann cells to provide a substrate for axonal infiltration, but also serves autocrine roles in inducing Schwann cell differentiation to a myelinating phenotype [64,65]. Greater laminin expression observed at 12-weeks in experimental groups may suggest a transition from the earlier time point, where Schwann cells have associated with regenerating axons, and have begun differentiating into myelinating phenotypes. Interestingly, contrary to temporal trends in collagen deposition observed in MARTEENI devices, Wurth et al. found significantly increased collagen deposition at 165 days compared to 28 days from intrafascicular polyimide electrodes [66]. The differences in trends between timepoints may be attributed to varying methods of implantation and resulting healing mechanisms, where SELINE electrodes were threaded directly into the sciatic nerve, versus MARTEENI devices implanted into the resected nerve. MARTEENI’s regenerative hydrogel may also provide benefits to chronic healing, however this determination would require an extended timepoint in future studies.

MARTEENI aims to integrate fields of neural and tissue engineering to engage with the nervous system while promoting regeneration in close proximity to the interface. Thus, a primary goal of the reported in vivo study was to assess nerve regeneration through MARTEENI devices in comparison to other clinically relevant therapeutic options, including the Isograft and Empty conduit. It is widely accepted that peripheral nerve tissue, with the addition of a guide, can successfully regenerate at sub-critical gap lengths [23]. For prosthetic-limb interfacing, it is not necessary to fabricate critical gap length devices for tissue-specific reinnervation; rather, it is more pertinent to ensure good reinnervation around the device and electrode. MARTEENI devices were implanted within a non-critical 5 mm gap; thus, it is hypothesized that MARTEENI devices should be fully regenerated upon excision at both time points. Additionally, it is not expected that the polyimide would impede regeneration, as other intrafascicular polyimide devices, such as the LIFE, FINE, and SELINE electrodes, show comparable axon densities to native tissue after implantation [13,66,67]. No statistical differences were observed in the axon densities between experimental groups and Fresh, except at the 12-week distal implant, where there was an increase in axon densities of MARTEENI and Empty groups (Fig. 7A). Additionally, axons in the distal implant appeared significantly smaller in diameter compared to Fresh. These observations could be attributed to axon sprouting of immature daughter axons in the distal implant [68]. Although higher axon densities were observed through MARTEENI devices, these smaller diameter sprouts (Fig. 7B) in the distal implant likely have not formed connections with the distal nerve stump and will die off before they can reach their functional target [68,69]. Thus, it is pertinent to conduct electrophysiological studies in the future to determine downstream reinnervation. Future work will include implantation of fully functional MARTEENIs to test the device efficacy, while simultaneously gathering longitudinal electrophysiological data. In the context of this study, MARTEENIs demonstrate equivalent and robust regeneration as compared to other clinical nerve guides; thus, we expect to be able to successfully measure neural signals in future studies.

Of note, quantitative analyses from this in vivo study show correlated relationships between several regenerative markers that can be linked to successful tissue regeneration. In Fig. 7C, a positive correlation between S100 and laminin expression is observed. These respective cell and ECM markers have been colocalized to one another in autocrine signaling pathways after injury and have been highly implicated in successful Schwann cell localization and subsequent nerve growth [64]. Additionally, our results also show a positive correlation between laminin expression and axon density. The basal lamina substrates that preferentially guide axons are largely composed of laminin [65]. These results suggest that MARTEENI devices do not disrupt regenerative pathways associated with successful regeneration.

Lastly, device artifacts, including undegraded hydrogel and fibrotic encapsulation of polyimide threads, were seen in MARTEENI devices. Although MARTEENI demonstrated high axon densities at 12 weeks, residual hydrogel remaining within the implant space limited some cross-sectional area that could be infiltrated by cells. Several studies have suggested that slow degradation of nerve guides may be advantageous to axonal regeneration [70–72]. Hsu et al. noted slow in vivo breakdown of polylactic acid conduits and equally steady increases in functional outputs to nearly complete recovery after 18 months [71]. Such noted studies have been primarily conducted with hollow conduit guides; thus, it is unclear how the degradation rate of intraluminal fillers affect long-term nerve growth. With these considerations, it may be advantageous to investigate hydrogel modifications that can render the scaffold more enzymatically tunable through the introduction of matrix metalloproteinase-degradable motifs [73–75].

In addition, fibrotic encapsulation of the MARTEENI polyimide threads was observed. Although polyimide is significantly less stiff than traditional silicon interfaces, it still possesses a modulus that is three orders of magnitude stiffer than neural tissue, and thus, this fibrotic response was unsurprising [27,66]. Capsules only appear around individual threads, indicating good spacing between sites to help prevent the convergence of one large foreign body. Importantly, the measured capsule thicknesses around MARTEENI threads show no statistical differences across implant regions or timepoints, suggesting that the capsule size has likely reached a steady-state and will not continue to grow. Concerns over increasing fibrotic capsule size include increased impedance requiring higher stimulation thresholds over time [76]. Wurth et al. previously demonstrated long-term functional stability of a polyimide-based intra-neural implant where capsule thicknesses of approximately 50 μm were measured at 4 weeks [66]. Moreover, MARTEENI showed capsule thicknesses comparable to the LIFE electrode, which show no alterations to CMAP amplitude through a chronic implant period [67]. Based on MARTEENI measured capsule thicknesses of approximately 20 μm at 12 weeks, it is expected that functional devices will still be able to record and stimulate single axons with high selectivity and low stimulation regimes over a chronic period. Future work will include investigation into substrate modifications including varying thread geometries and conductive coatings [77,78] to aid in reducing the foreign body response and its limiting effects.

5. Conclusions

Limitations from existing prosthetic-limb neural-interfacing technology, including stiffness mismatch, low axonal population sampling, and long-term signal decay, present an inherent need for innovative tissue-engineered approaches. Our work establishes the feasibility of a new multidisciplinary approach, “MARTEENI,” that combines a tissue-engineered electronic nerve-interface technology with magnetically templated scaffolds.

The work presented has demonstrated the utility of magnetic templating to fabricate hydrogel scaffolds with mechanical and structural properties that mimic native nerve tissue. In addition, it is shown that 3D scaffolds with stiffnesses and microarchitecture resembling native nerve promote better Schwann cell migration in vitro. Based on mechanical and in vitro cellular observations, we selected a templated hydrogel composition for further integration into MARTEENI devices. Polyimide threads were shown to be supported within the chosen hydrogel composition and did not disrupt the templating process. A pilot in vivo experiment was conducted to evaluate general tissue remodeling and to ensure no adverse foreign-body effects through MARTEENI devices implanted within a 5-mm-long rat sciatic-nerve transection model. Overall, MARTEENI promotes cellular infiltration and ECM deposition that support successful nerve regeneration. High axon densities were observed through the implant; however, it remains unclear whether these axons have reinnervated their target from histological analysis alone.

MARTEENI devices did not cause severe adverse effects to the regenerated tissue. However, there still appeared to be undegraded hydrogel remaining in the implant space after 12 weeks and foreign-body formations around individual polyimide threads. Future work will include the development of alternative hydrogel chemistries with accelerated degradation properties to facilitate greater axonal infiltration. Fibrotic capsule thickness did not appear to increase between the early and late time points; however, it is unclear how impedance and the signal-to-noise ratio will change over time due to the foreign body capsules. Thus, future studies will be focused on implantation of functional MARTEENI devices to assess neural functional recovery and MARTEENI efficacy to record and stimulate action potentials. Further, expansion of the device technology will include exploiting the hydrogel to support multiple thread layers in the interface through the entire tissue geometry.